Main findings and their interpretation

6.1.1 hCG-h, %hCG-ŚĂŶĚĨƌĞĞŚ'ɴ

We found that prior pre-eclampsia and lower serum %hCG-h MoM in the first trimester were associated with late-onset and non-severe pre-eclampsia, whereas lower serum PlGF MoM in the first trimester was associated with early-onset and severe pre-eclampsia among high-risk women of the PREDO cohort. Of the biomarkers, free hCGɴ MoM was higher in women who developed severe pre-eclampsia than in women who did not develop severe pre-eclampsia, whereas low serum PAPP-A MoM was associated with non-severe pre-eclampsia. Despite these associations, multivariate models constructed with regularised logistic regression produced only modest prediction rates for all pre-eclampsia and its subtypes in the PREDO cohort consisting of women with a priori risk factors for pre-eclampsia. Our results suggest that %hCG-h does not add any value to pre-eclampsia prediction over previously tested variables using a multivariate approach, at least not in a high-risk population and in the late first trimester.

The role of hCG-h in the pathophysiology of pre-eclampsia is yet to be elucidated, but evidence from the in vitro and in vivo study of Guiboudenche and colleagues showed that extravillous cytotrophoblasts initiate the production of hCG-h during their differentiation from proliferative cytotrophoblasts to invasive cytotrophoblasts in normal pregnancies, and therefore the hCG-h circulating in maternal serum may reflect the invasion process of trophoblasts during the first trimester (9).

Our result is inconsistent with the result of Keikkala and colleagues (8), because their case–control study on a screening cohort showed that lower serum %hCG-h in the first trimester was associated with subsequent early-onset pre-eclampsia. There are two factors that may explain the contradiction: 1) In the study of Keikkala and colleagues the study cohort comprised an unselected screening population, whereas our cohort comprised prospectively recruited high-risk women. 2) The median sampling time in the study of Keikkala and colleagues was three weeks earlier (about 10 weeks of gestation) than in our study (about 13 weeks of gestation). Since the concentration of hCG-h and especially %hCG-h decreases rapidly (258) during early pregnancy, it is possible that this time interval is significant regarding the serum %hCG-h evolvement, because total hCG reaches its peak value at 10–11 weeks of gestation and decreases thereafter (21, 368).

There are no studies regarding how hCG-h or %hCG-h develops in early- or late-onset pre-eclampsia. Although, classically there is an idea of late-onset pre-eclampsia being more of a ‘maternal disease’ and early-onset pre-eclampsia being a ‘placental disease’, there may be milder and later-developing defects of the placenta associated with late-onset pre-eclampsia which may affect the evolvement of serum hCG-h concentration.

Soto and colleagues showed in their work on placentas delivered from pregnancies complicated with onset pre-eclampsia that about half of the patients with late-onset pre-eclampsia had placental lesions consistent with abnormally low perfusion of the placenta, while only 2% of the placentas delivered from normal pregnancies had

the same kind of lesions (369). Furthermore, the regression line reflecting hCG-h concentrations against gestational weeks in the scatterplot showed lower values in early-onset pre-eclampsia before 12 weeks gestation than in the other groups. This agrees with previous publications (8, 261), but in the present study only nine women had early-onset pre-eclampsia. Thus, it is not possible to draw any definitive conclusions. Our result with respect to hCG-h and %hCG-h was in accordance with the study of Brennan and colleagues, since they found a possible association with low urine hCG-h after 10 weeks of gestation and late-onset pre-eclampsia. Although the difference between women who subsequently developed late-onset pre-eclampsia and controls was significant, the number of women affected was only five (263). In a case–

control study consisting of women participating in a first trimester screening of chromosomal abnormalities, there was no difference in serum %hCG-h concentration between women who did and did not develop pre-eclampsia when the samples were collected at 14–17 weeks of gestation (264).

In univariate analyses, the concentrations of studied biomarkers showed specific associations to different phenotypes of pre-eclampsia. The women who developed severe pre-eclampsia had higher free hCGɴ MoM than women who did not develop severe pre-eclampsia. Previous studies have reported decreased first trimester concentrations of free hCGɴ in women who subsequently developed pre-eclampsia compared to women who did not develop pre-eclampsia (370-372), but in substantially more studies there has not been any association between free hCGɴ and pre-eclampsia (311, 333, 373-376). Furthermore, in the systematic review of Kuc and colleagues, the free hCGɴ did not show any association to pre-eclampsia (375). A study of Di Lorenzo and colleagues reported that in their first trimester screening cohort higher free hCGɴ concentrations were associated with early-onset pre-eclampsia (377).

High maternal serum free hCGɴ MoM seen in trisomy 21 (Down syndrome) pregnancies (median 1.8 MoM in the study of Macintosh and colleagues) is thought to be related to delayed maturation of the placenta (378). However, in our study, although women who developed severe pre-eclampsia had higher median free hCGɴ MoM than those who did not develop severe pre-eclampsia, the median free hCGɴ MoM was below 1.0, and the women who did not develop severe pre-eclampsia had even lower median free hCGɴ MoM. The median free hCGɴ MoM was especially low in women who subsequently developed late-onset and non-severe pre-eclampsia, when compared to women who did not develop these conditions. The difference was not significant, but it may have had effect on the median free hCGɴ MoM value of women who did not develop severe pre-eclampsia.

6.1.2 PAPP-A and PlGF

The maternal serum PAPP-A MoM was lower in women who developed pre-eclampsia compared to women who did not develop pre-eclampsia, but the difference was not significant. However, lower maternal serum PAPP-A MoM had a significant association with non-severe pre-eclampsia. Women with early-onset pre-eclampsia had also similarly decreased PAPP-A MoM, but the difference between women who did and did not develop early-onset pre-eclampsia was not significant. Evidence from previous studies shows that there is an association between low concentrations of first trimester serum PAPP-A and pre-eclampsia (371, 373, 374, 379-382). As part of a prediction

model, serum PAPP-A is better at predicting early-onset and pre-term pre-eclampsia than late-onset or term pre-eclampsia in the first trimester (188, 194).

Serum PlGF concentration showed an association with early-onset and severe pre-eclampsia in the first trimester, especially when combined with maternal factors and other variables. This is in accordance with previous studies (188, 345).

6.1.3 Multivariate model

Comparison of our multivariate model to previous publications is not possible since there are no other studies conducted in the same kind of a cohort. It would have been easier to compare our results to, for example, results of studies conducted with the FMF algorithm, if we knew the exact prediction rates of the inclusion criteria that were chosen for the PREDO cohort. In our study, results showed how maternal factors predict eclampsia among women that already have increased risk for pre-eclampsia. Hence, we should add the a priori risk to those prediction rates that were calculated from the multivariate models constructed with logistic regression to make our results comparable with previous studies conducted in first-trimester screening studies. With the NICE recommendation for screening pre-eclampsia with maternal factors, detection rates were 34%, 39% and 41% for pre-term (delivery <37 weeks of ŐĞƐƚĂƚŝŽŶͿ͕ƚĞƌŵ;ĚĞůŝǀĞƌLJшϯϳǁĞĞŬƐŽĨŐĞƐƚĂƚŝŽŶͿĂŶĚĞĂƌůLJpre-eclampsia (delivery <32 weeks of gestation) with a screen positive rate of 10.2%, respectively (Table 9) (186).

There are two obstacles that make comparison difficult: the categories of pre-eclampsia by gestational weeks are different when compared to our study, and the maternal factors with which women are included to the high-risk group are not exactly the same (Table 9 and Table 15). However, prediction performance of the NICE recommendation may provide a clue as to what the a priori risk of our cohort might be. Nevertheless, it was not possible to calculate the prediction rates of inclusion criteria in our study (i.e., positive likelihood ratio of a priori risk) due to the nature of our cohort (selection made using maternal factors, no data on the unselected population).

The detection rates achieved with the multivariate approach (Bayes theorem) in first-trimester screening cohorts are 90% for early-onset pre-eclampsia and 60% for late-onset pre-eclampsia with a 10% screen positive rate (188), whereas the detection rate (sensitivity) in our study for early-onset eclampsia was 20% and for late-onset pre-eclampsia 32% with a 10% false positive rate. Although the statistical methods between the study of Tan and colleagues (188) and our study were different, one could conclude that prediction of pre-eclampsia with a multivariate model combining maternal factors, MAP, biomarkers and Uta-PI succeeds better in a screening cohort than in a high-risk cohort.

6.1.4 PlGF and LDA

The main finding of our study was that high-risk women who had LDA treatment (100 mg/d), started before 14 weeks of gestation for prevention of pre-eclampsia, had higher concentrations of serum PlGF than high-risk women who had placebo treatment. This difference was evident from mid-gestation onwards. There is one study investigating the effect of LDA on serum PlGF concentrations during pregnancy that was published before our study (298). Powers and colleagues could not find a difference in

maternal serum PlGF concentrations between those high-risk women who used LDA and those who used placebo during pregnancy. Since publication of our study, two additional studies have been published on the subject. Mone and colleagues came to the same conclusion as Powers and colleagues (299). Their cohort consisted of low-risk nulliparous women. However, in both studies the LDA dose was lower (60 mg/d) than in our study (100 mg/d). In the study of Mayer-Pickel and colleagues, it was found that aspirin may have a favourable effect on the sFlt-1/PlGF in women with a pathologic first-trimester screening for pre-eclampsia, with the most prominent effects at a dosage of 150 mg/d, and the effect appeared to be dose-dependent (300). Their finding is in accordance with our finding and with a systematic review and meta-analyses by Roberge and colleagues, in which it was concluded that the dose of 100 mg/day seems to be the minimum dose at which the effect of aspirin becomes evident, and the effect of aspirin for the prevention of pre-eclampsia, severe pre-eclampsia and foetal growth restriction is dose-dependent (383). Our result along with the result of Mayer-Pickel and colleagues supports the preventive use of LDA.

Another finding in our study was that the increase in PlGF concentration was lower in the medium-risk women with pre-eclampsia without any treatment as well as in high-risk women who developed pre-eclampsia and had placebo treatment compared to the other groups. This is in line with earlier studies that have shown that serum PlGF concentration is reduced in women who subsequently develop pre-eclampsia already in early pregnancy when compared to women with normal pregnancies (384), and as in our study, this difference becomes clearer with advancing pregnancy (11, 385, 386).

6.1.5 Pre-eclampsia and haem scavenger proteins in the late second trimester We found that there are differences in plasma concentrations of haem scavenger proteins Hpx and A1M between HRW, HRPE and LRW at the end of the second trimester. Plasma A1M appeared to be higher in HRW when compared to LRW or HRPE at 26–28 weeks of gestation.

All previous studies on Hb scavengers and pre-eclampsia are conducted either in the first trimester or within 24 hours before delivery. Studies conducted in the first trimester indicated increased concentrations of Hb, HbF/Hb (ratio) and A1M, as well as reduced concentrations of Hpx (17, 19). In studies where blood samples were drawn within 24 hours before delivery, both HbF and adult Hb concentrations, as well as haem and A1M, were increased and Hpx and Hpx activity, as well as HO-1 concentration, were decreased in women who had pre-eclampsia (17, 19). There are no previous studies on maternal A1M concentrations in the late second trimester. We hypothesised that the differences seen in previous studies near delivery between pre-eclamptic and normotensive women would be detectable during the late second trimester. However, the median concentrations of Hpx did not differ between the groups after we had excluded controls and women who used LDA, and surprisingly, HRW had higher median A1M concentrations than HRPE and LRW. Contradiction between our hypothesis, which was based on previous studies at different time points during pregnancy, and our findings may be that in previous studies normotensive pregnant women had been studied as a single group, irrespective of their risk factors for pre-eclampsia. The elevated A1M concentration in HRW may be explained by activation of an endogenous protection system against oxidative stress, since A1M is up-regulated during oxidative

stress in general and by haemolysis (349). It is tempting to speculate that these mechanisms may function better in high-risk women with pregnancies not complicated by pre-eclampsia.

6.1.6 Changes in plasma concentrations of Hpx and A1M

We hypothesised that the Hpx concentrations measured at three different time points during pregnancy would be lower in HRPE compared to the other groups due to a pre-eclampsia–induced increase in HbF concentration in the maternal circulation and by subsequent activation of the Hb scavenger system. However, the plasma Hpx concentration in HRPE did not differ from the plasma Hpx concentration of LRW, and it was higher compared to the HRW during the first half of pregnancy, suggesting a protective role for low plasma Hpx concentration in HRW.

There is some contradiction between our study results and those of Anderson and colleagues who found that plasma collected in the first trimester had lower Hpx concentrations among women who subsequently developed early-onset pre-eclampsia compared to women who did not develop pre-eclampsia (19). This may be explained by the multifactorial mechanisms affecting the concentration and activity of Hpx.

Changes in the plasma Hpx concentration reflect the balance between hepatic synthesis of Hpx, decreases in Hpx due to haem scavenging and rate of Hpx degradation.

One theory that could explain our unexpected result is that the expression and activity of Hpx may be affected by systemic inflammation and oxidative stress. A study conducted on patients with type 1 diabetes mellitus showed that the expression of Hpx can be induced through an ROS-dependent mechanism (387). Similarly, the expression of Hpx might be induced due to ROS-mediated oxidative stress in high-risk women destined to develop pre-eclampsia, which would explain the increased level of Hpx in pre-eclampsia as shown in our study in HRPE, where obesity, chronic hypertension and diabetes mellitus were among the inclusion criteria for recruitment to the high-risk cohort. These conditions may predispose women to inducement of ROS-mediated oxidative stress (388-390).

We found that BMI had an association with Hpx concentration, which could explain some of the differences between the study groups, although most of the differences remained significant after adjustment of the data for BMI. We hypothesised that the adjustment of Hpx concentrations for BMI might better reflect the role of the placenta in the results. However, it is also known that obesity is an independent risk factor for pre-eclampsia (391).

We found that HRW had a unique profile of haem scavenger proteins during pregnancy.

Firstly, unlike in HRPE and LRW, their plasma concentration of Hpx did not change during the study period and it was lower during the first half of the pregnancy when compared to the other two groups. Secondly, the A1M concentration increased during the first half of the pregnancy and stayed at the higher level thereafter when compared to the other two groups. The change during the first half of the pregnancy was opposite to the change seen in HRPE, while there was no change of plasma A1M concentration in LRW. This phenomenon may be associated with a reduced risk of pre-eclampsia despite clinical risk factors. However, further information about the mechanisms

behind our finding could have been gained by investigating oxidative stress markers at the same time with Hb scavengers.

One of our findings is consistent with previous studies: A1M concentration was higher in the first trimester in HRPE. Recent studies have suggested that A1M is a promising first-trimester biomarker for prediction of pre-eclampsia (17, 19). The higher A1M concentration of HRW already seen in Study III was found to be evident from mid-gestation onwards, and the speculation about its cause is the same.

The plasma Hpx concentration started to decrease from the first trimester in PESGA, as we had hypothesised, but the difference in concentrations and in changes of concentrations were not statistically significant. However, the number of women in this group was rather small, thus the risk of type II error in this situation is possible. When these findings are reflected to the present knowledge of the Hb scavenger system, it could be anticipated that concentration of free HbF is higher in the foeto-placental circulation in PESGA than in normal pregnancies (392). Thus, it is possible that HbF and increased ROS may damage the placenta–blood barrier causing leakage of HbF into the maternal circulation, which in turn leads to Hpx depletion (223).

We also found that the concentration of A1M in PESGA remains high during the first half of pregnancy, but in PEAGA it decreases. In HRPE, high A1M concentration at and after mid-gestation is associated with SGA.

Several factors may have an influence on the concentration and changes in concentration of Hb scavenger proteins. Firstly, we found that BMI is associated with Hpx concentration. Secondly, obesity may also indirectly affect the concentration of Hpx and A1M since it predisposes to a pro-inflammatory state and oxidative stress (393), conditions that occur in normal pregnancy and are exacerbated in pre-eclampsia (151). Thirdly, the severity of placentation defect and its consequences to placental circulation and function may have crucial contributions to the gene expression of HbF in both the foetus and placenta, as well as to the integrity of the placenta–blood barrier (394). Therefore, plasma concentrations of Hpx or A1M in different subgroups of pregnant women may reflect different combinations of these factors. Thus, the relation of the haem scavenger system to pre-eclampsia should be assessed according to the pathophysiology of the disorder since, in light of present evidence, there are several mechanisms that may lead to the same clinical disorder (121, 395). Further studies should assess maternal cardiac function simultaneously with the concentration of haem and markers of oxidative stress and inflammation to further clarify whether the suggested mechanism presented in our study has clinical relevance.